chacha/asm/chacha-x86[_64].pl: fix typos and logical errors.

Thanks to: David Benjamin of Chromuim.

RT#4305

Reviewed-by: Rich Salz <rsalz@openssl.org>

diff --git a/crypto/chacha/asm/chacha-x86.pl b/crypto/chacha/asm/chacha-x86.pl
index 60d6048..e2019aa 100755 (executable)
--- a/crypto/chacha/asm/chacha-x86.pl
+++ b/crypto/chacha/asm/chacha-x86.pl
@@ -219,7 +219,7 @@ if ($xmm) {
        &dec    ($b);
        &jnz    (&label("loop"));
 
        &dec    ($b);
        &jnz    (&label("loop"));
 

-       &mov    ($b,&wparam(3));                # load len
+       &mov    ($b,&wparam(2));                # load len

 
        &add    ($a,0x61707865);                # accumulate key material
        &add    ($b_,&DWP(64+4*4,"esp"));
 
        &add    ($a,0x61707865);                # accumulate key material
        &add    ($b_,&DWP(64+4*4,"esp"));

diff --git a/crypto/chacha/asm/chacha-x86_64.pl b/crypto/chacha/asm/chacha-x86_64.pl
index 41dbef5..1a87cf8 100755 (executable)
--- a/crypto/chacha/asm/chacha-x86_64.pl
+++ b/crypto/chacha/asm/chacha-x86_64.pl
@@ -344,8 +344,8 @@ $code.=<<___;
 .align 16
 .Ltail:
        mov     @x[0],4*0(%rsp)
 .align 16
 .Ltail:
        mov     @x[0],4*0(%rsp)

-       xor     %rbx,%rbx

        mov     @x[1],4*1(%rsp)
        mov     @x[1],4*1(%rsp)

+       xor     %rbx,%rbx

        mov     @x[2],4*2(%rsp)
        mov     @x[3],4*3(%rsp)
        mov     @x[4],4*4(%rsp)
        mov     @x[2],4*2(%rsp)
        mov     @x[3],4*3(%rsp)
        mov     @x[4],4*4(%rsp)


@@ -521,11 +521,11 @@ $code.=<<___;
 
 .Loop_tail_ssse3:
        movzb   ($inp,%rbx),%eax
 
 .Loop_tail_ssse3:
        movzb   ($inp,%rbx),%eax

-       movzb   (%rsp,%rbx),%edx
+       movzb   (%rsp,%rbx),%ecx

        lea     1(%rbx),%rbx
        lea     1(%rbx),%rbx

-       xor     %edx,%eax
+       xor     %ecx,%eax

        mov     %al,-1($out,%rbx)
        mov     %al,-1($out,%rbx)

-       inc     %rbp
+       dec     $len

        jnz     .Loop_tail_ssse3
 
 .Ldone_ssse3:
        jnz     .Loop_tail_ssse3
 
 .Ldone_ssse3: